Mtor-Specific Inhibitors in BRAF-Mutant Melanoma Cells Counteracted by Dual Pathway Blockade

www.impactjournals.com/oncotarget/ Oncotarget, Vol. 7, No. 4

Primary cross-resistance to BRAFV600E-, MEK1/2- and PI3K/ mTOR-specific inhibitors in BRAF-mutant melanoma cells counteracted by dual pathway blockade

Ilaria Penna1,2, Alessandra Molla1,2, Giulia Grazia1,2, Loredana Cleris2, Gabriella Nicolini1,2, Federica Perrone3, Benedetta Picciani3, Michele Del Vecchio4, Filippo de Braud4, Roberta Mortarini1,2,*, Andrea Anichini1,2,* 1Human Tumors Immunobiology Unit, Fondazione IRCCS Istituto Nazionale dei Tumori, Milan, Italy 2Department of Experimental Oncology and Molecular Medicine, Fondazione IRCCS Istituto Nazionale dei Tumori, Milan, Italy 3Laboratory of Molecular Pathology, Department of Pathology and Laboratory Medicine, Fondazione IRCCS Istituto Nazionale dei Tumori, Milan, Italy 4Medical Oncology Unit 1, Department of Medical Oncology, Fondazione IRCCS Istituto Nazionale dei Tumori, Milan, Italy *These authors contributed equally to this work Correspondence to: Andrea Anichini, e-mail: [email protected] Keywords: melanoma, BRAF, MEK, PI3K/mTOR, apoptosis Received: June 30, 2015 Accepted: November 22, 2015 Published: December 14, 2015

ABSTRACT Intrinsic cross-resistance to inhibition of different signaling pathways may hamper development of combinatorial treatments in melanoma, but the relative frequency of this phenotype and the strategies to overcome this hurdle remain poorly understood. Among 49 BRAF-mutant melanoma cell lines from patients not previously treated with target therapy, 21 (42.9%) showed strong primary resistance

(IC50 > 1 μM) to a BRAFV600E inhibitor. Most of the BRAF-inhibitor-resistant cell lines showed also strong or intermediate cross-resistance to MEK1/2- and to PI3K/ mTOR-specific inhibitors. Primary cross-resistance was confirmed in an independent set of 23 BRAF-mutant short-term melanoma cell cultures. MEK1/2 and PI3K/mTOR co-targeting was the most effective approach, compared to BRAF and PI3K/mTOR dual blockade, to counteract primary resistance to BRAF inhibition and the cross- resistant phenotype. This was shown by extensive drug interaction analysis, tumor growth inhibition assays in-vivo, p-ERK and p-AKT inhibition, promotion of melanoma apoptosis, apoptosis-related protein modulation, activation of effector caspases and selective modulation of genes involved in melanoma drug resistance and belonging to the ERK/MAPK and PI3K/AKT canonical pathways. Compared to co-targeting of mutant BRAF and PI3K/mTOR, the association of a MEK1/2 and a PI3K/mTOR inhibitor was more effective in the activation of Bax and of caspase-3 and in the induction of caspase-dependent melanoma apoptosis. Furthermore Bax silencing reduced the latter effects. These results suggest that intrinsic resistance to BRAF inhibition is frequently associated with primary cross-resistance to MEK and PI3K/ mTOR blockade in BRAF-mutant melanoma and provide pre-clinical evidence for a combinatorial approach to counteract this phenotype.

www.impactjournals.com/oncotarget 3947 Oncotarget INTRODUCTION inhibitors. Extensive drug interaction analysis on all 49 cell lines and mechanistic studies in cross-resistant cell lines Inhibitors of BRAFV600E have changed the clinical indicated that co-targeting of MEK1/2 and PI3K/mTOR, management of patients with BRAF-mutant advanced in-vitro and in-vivo, was a more effective combinatorial melanoma, since significant improvements in progression- treatment, compared to co-targeting of BRAF and PI3K/ free survival (PFS) and in overall survival (OS) have been mTOR, to counteract the primary cross-resistant phenotype. reported in Phase 3 trials [1–2]. These inhibitors can induce objective responses or stabilization of disease in RESULTS a high fraction of patients [2–3], although relapse occurs due to adaptive [4] or acquired [5] resistance mechanisms. Primary resistance to BRAFV600E inhibition In addition, ~20% of patients show primary/intrinsic is associated with cross-resistance to MEK1/2 resistance to BRAF inhibitors and experience tumor and PI3K/mTOR inhibitors in BRAF-mutant progression at first assessment during therapy [2–3]. melanoma cells Lack of response has been observed even in trials with MEK inhibitors as Trametinib and Selumetinib/AZD6244 We used 49 BRAF-mutant melanoma cell lines [6–7], pointing to primary resistance also to the targeting isolated from surgical specimens of patients not previously of MEK. To counteract resistance to BRAF inhibitors, treated with BRAFV600E inhibitors, nor with any several clinical trials based on combinatorial targeting other target-specific inhibitor, to test responsiveness to of BRAF and MEK have been carried out recently, and BRAFV600E (PLX4720), MEK1/2 (AZD6244), dual results have indicated a significant improvement in PFS PI3K/mTOR (BEZ235) and dual mTORC1/2 (AZD8055) and (in one study) in OS, compared to monotherapy with inhibitors (Figure 1A). Three susceptibility groups were

the BRAF inhibitor alone [8–10]. However, strikingly, in defined by ranking cell lines based on their IC50 values for

all these studies ~30% of the patients showed progressive PLX4720. Strong resistance (IC50 > 1 μM) was found in disease at 6 months of treatment [8–10], pointing to the 21/49 cell lines (42.9%, group 1), intermediate resistance

potential role of cross-resistance mechanisms. Indeed, (IC50 = 0.1 to 1 μM) and susceptibility (IC50 < 0.1 μM) to primary cross-resistance to BRAF and MEK inhibitors has PLX4720 were observed in 16/49 (32.7%, group 2) and been documented in a subset of melanomas, where it is 12/49 cell lines (24.5%, group 3), respectively (Figure 1A related to the MITF profile [11–13], and in cell lines [14]. and Supplementary Table 1D for descriptive statistics). In the attempt to address all these limitations Seventeen out of 21 melanoma lines (marked with an associated with targeting of a single (MAPK) pathway, “x” in Figure 1A) classified in group 1, showed either several pre-clinical studies have suggested that combi strong or intermediate primary resistance to MEK1/2 and natorial targeting of MAPK and PI3K/mTOR intracellular PI3K/mTOR inhibitors. Strong or intermediate cross- pathways may be a potential approach to improve target resistance to PI3K/mTOR inhibitors was also found in therapy of melanoma and to overcome resistance and 11/16 cell lines in group 2, while only two cell lines in cross-resistance mechanisms [14–17]. However, key group 3 (PLX4720 susceptible) showed strong resistance questions to be answered are whether BRAF-mutated to the PI3K/mTOR inhibitors (Figure 1A). Hierarchical

melanomas, with intrinsic resistance to BRAF and/or clustering of log-transformed and normalized IC50 values MEK inhibitors, also have primary cross-resistance to confirmed the existence of distinct subsets of cell lines PI3K/mTOR inhibitors and which is the frequency of such with cross-resistance to all inhibitors or susceptible to cross-resistant phenotype. A second issue to be addressed all of them (Figure 2A). Spearman correlation analysis

is whether combinatorial targeting of different signaling of IC50 values for all six possible combinations of four pathways is effective even in the tumors with primary inhibitors showed that all the susceptibility profiles were cross-resistance. significantly correlated (Figure 2B). Recently, we observed instances of primary cross- To validate these findings, an independent panel of resistance to a MEK1/2 (AZD6244) and to a dual PI3K/ short-term melanoma cell cultures (all tested between the mTOR (BEZ235) inhibitor in some melanoma cell lines third and fifth in-vitro passage), obtained from 23 BRAF- (18). Starting from this initial evidence, in this study we mutant metastatic specimens of patients not previously assessed responsiveness to BRAFV600E-, MEK1/2-, treated with target-specific inhibitors, was used to test dual PI3K/mTOR- and dual mTORC1/2-specific inhibitors responsiveness to the same set of inhibitors. The same in 49 BRAF-mutant melanoma cell lines and in an classification into three subsets based on ranking of

independent set of 23 BRAF-mutant short-term melanoma PLX4720 IC50 values was applied. We found that 6/6 cell cultures, all isolated from patients never treated with PLX4720-resistant melanoma cell cultures (group 1)

target therapy. Among cell lines with strong intrinsic showed strong (i.e. IC50 > 1 μM) or intermediate

resistance to the BRAF inhibitor (IC50 > 1 μM, n = 21), (i.e. IC50 > 0.1 μM) cross-resistance to MEK1/2 and 81% (seventeen) showed strong or intermediate cross- PI3K/mTOR inhibitors, and 11/13 cultures in group 2 resistance to the MEK1/2- and the PI3K/mTOR-specific (intermediate resistance to PLX4720) showed also strong www.impactjournals.com/oncotarget 3948 Oncotarget Figure 1: Responsiveness to BRAF-V600E-, MEK1/2- or PI3K/mTOR-specific inhibitors in BRAF-mutant melanoma cell lines. (A) Susceptibility to PLX4720 (BRAFV600E inhibitor), AZD6244 (MEK1/2 inhibitor), BEZ235 (dual PI3K/mTOR inhibitor)

and AZD8055 (dual mTORC1/2 inhibitor), shown as IC50 values (μM), was assessed by a 72 h MTT assay in a panel of 49 melanoma

cell lines. IC50 values obtained through non linear regression analysis of twelve-point dose-response curves spanning 4 logs of inhibitor

concentrations. Melanomas were distinguished into three groups after ranking based on PLX4720 IC50 values (group 1: IC50 > 1 μM; group

2: IC50 > 0.1 μM but < 1 μM; group 3: IC50 ≤ 0.1 μM). (B) Color code used for highlighting differences in IC50 values in panel A. www.impactjournals.com/oncotarget 3949 Oncotarget or intermediate cross-resistance to PI3K/mTOR inhibitors cross-susceptible group 3 (Supplementary Figure 1A), (Figure 3A). As a control, 10 short-term melanoma cell indicated a strong suppression of melanoma growth by cultures from tumors with wt BRAF were characterized AZD6244, PLX4720, BEZ235 and AZD8055, often for responsiveness to the four inhibitors. As expected [19], detected at the lowest inhibitor dose (0.1 μM). In contrast, all the BRAF wt melanoma cell cultures were strongly clonogenic assays on representative cell lines (Me35, Me6, resistant to PLX4720, but some of them also showed Me13) and short-term melanoma cell cultures (Me_cc102) strong resistance to the MEK1/2 or to the PI3K/mTOR from group 1 (Supplementary Figure 1B) showed a partial inhibitors (Figure 3B). Interestingly, the melanoma cell or markedly reduced inhibitory effect by AZD6244 (on culture Me_cc135, with intermediate cross-resistance, was Me35 and Me_cc102), by PLX4720 (on Me35, Me6, Me13 isolated from a specimen of a patient who subsequently and Me_cc102), and by AZD8055 (on Me35, Me13 and (4.4 months after Me_cc135 isolation) was treated with a Me_cc102). BEZ235 exerted a reduced inhibitory effect BRAF inhibitor and underwent progressive disease after on Me35, even at the highest dose, in agreement with the

two cycles of therapy. In contrast, melanoma cell cultures high IC50 value in this cell line (Supplementary Figure 1B). Me_cc111 and Me_cc128, with a cross-susceptible Taken together, these assays confirmed that cell lines and phenotype, were isolated from patients who subsequently short-term melanoma cell cultures in group 1 showed (75.4 and 2.8 months, after Me_cc111 and Me_cc128 markedly reduced responsiveness to multiple inhibitors. isolation, respectively) were treated with the association The panel of 49 melanoma cell lines shown in of a BRAF and a MEK inhibitor or in monotherapy with Figure 1, was further characterized for several molecular or a MEK inhibitor and experienced a partial response or a phenotypic features associated with drug resistance [20–23], complete response, respectively. but no significant association was found, between the drug Twelve days clonogenic assays on representative susceptibility groups and: a) the PTEN, MDM4 and MDM2 cell lines (Me43 and Me71) and short-term melanoma expression levels; b) the constitutive p-ERK, p-AKT and cell cultures (Me_cc117 and Me_cc128) from the p-S6 levels (Supplementary Table 1A–1C and 1E–1G).

Figure 2: Responsiveness profiles of BRAF-mutant human melanoma cell lines to BRAFV600E-, MEK1/2-, and PI3K/

mTOR-specific inhibitors are significantly correlated. (A) Hierarchical clustering of normalized and Log-transformed IC50 values

for the indicated inhibitors in 49 melanoma cell lines. Each IC50 value (as listed in Figure 1) was normalized against the median IC50 value

of each inhibitor in the whole panel. (B) Spearman correlation analysis of IC50 values for each of the six possible combinations of the four inhibitors in the whole panel of melanoma cell lines. www.impactjournals.com/oncotarget 3950 Oncotarget Figure 3: Responsiveness to BRAF-V600E-, MEK1/2- or PI3K/mTOR-specific inhibitors in short-term melanoma cell

cultures. (A, B) Susceptibility to PLX4720, AZD6244, BEZ235 and AZD8055, shown as IC50 values (μM), was assessed as described in the legend to Figure 1, in a panel of 33 melanoma cell cultures (Me_cc) bearing mutant BRAF (n = 23, A) or wt BRAF (n = 10, B). Melanoma cell cultures were established from surgical specimens of lymph node metastases and were tested between the third and fifth in-vitro passage. Short term melanoma cell cultures from BRAF-mutant lesions were distinguished into three groups after ranking based

on PLX4720 IC50 values as in Figure 1. BRAF genotype: mut: BRAFV600E; mut*: BRAFV600K WT. BRAF wild type. IC50 values were highlighted by the color code indicated in Figure 1B. www.impactjournals.com/oncotarget 3951 Oncotarget We also assessed the MITF phenotype of the cell lines and by the Chou-Talalay method [24]. To this end, 49 cell short-term melanoma cell cultures, as either high or low lines belonging to the three susceptibility groups were expression of this transcription factor has been associated treated with AZD6244-BEZ235, or PLX4720-BEZ235 or with drug resistance in melanoma [11–13]. We found that AZD6244-AZD8055 associations. For each association melanoma cell lines retained the MITF phenotype of the of inhibitors, twelve different combinations of doses corresponding lesions, but both MITFhi and MITFlo cell (indicated at the top of Figure 4A) were evaluated, lines and short term cultures were found in each of the yielding a 1,764-point drug interaction matrix (49 cell three susceptibility groups (data not shown). lines by 36 drug combinations). Strong synergistic Taken together, these results indicated that intrinsic effects, documented by CI values < 0.3 (Figure 4A and resistance to BRAFV600E inhibition can be frequently 4B for color codes and meaning of CI values and associated with cross-resistance to MEK1/2 and/or PI3K/ Supplementary Table 2 for primary data), were achieved mTOR inhibitors in BRAF-mutant melanoma cells. on almost all cell lines from group 2 and 3. Interestingly, strong synergism was observed at most drug dosing Co-targeting of MAPK and PI3K/mTOR combinations, against 20/21 melanoma cell lines in pathways in melanoma cells with a primary group 1 (Figure 4A), including 16/17 lines with the cross- cross-resistant phenotype has synergistic effects resistant phenotype (Figure 4A, arrows). By clonogenic in vitro and anti-tumor activity in vivo assays (Supplementary Figure 2A, 2B) AZD6244-BEZ235 and PLX4720-BEZ235 combinatorial treatments exerted a To test whether combinatorial targeting of MAPK strong inhibitory effect on melanoma growth not only on and PI3K/mTOR pathways could overcome primary two cell lines from group 3 (Me43 and Me71), but also on cross-resistance, we performed drug interaction analysis three cell lines from group 1 (Me35, Me6, Me13).

Figure 4: Synergistic drug interaction by co-targeting of MAPK and PI3K/mTOR pathways in melanoma cell lines. (A) Drug interaction analysis by the association of AZD6244 and BEZ235, PLX4720 and BEZ235, or AZD6244 and AZD8055 was carried out by the Chou-Talalay method in three groups of melanoma cell lines with different responsiveness to PLX4720 (Group 1, Group 2 and Group 3 as defined in Figure 1). Results for each of the indicated combinations of inhibitors (drug doses listed at the top of the Figure), for each cell line, are shown as Combination Indexes (CI) values by a color code shown in panel (B) Arrows: cell lines (n = 17) with a strong cross-resistant phenotype as in Figure 1. Blue lines: Wilcoxon matched pair test of CI values observed by AZD6244-BEZ235 vs. PLX4720- BEZ235 combinations having equivalent drug dosing. **p < 0.01. (B) color code for CI values, range of values and corresponding meaning (antagonism, additivity, synergy) according to ref. 24. Red indicates antagonism, green indicates synergism. www.impactjournals.com/oncotarget 3952 Oncotarget In addition, the AZD6244-BEZ235 combination dose compared to AZD6244 (see legend to Figure 5). In induced a significantly stronger synergistic effect (lower these normalized conditions, the AZD6244-BEZ235 and CI values) against the whole panel of 49 melanoma cell AZD6244-AZD8055 associations were more effective lines compared to PLX4720-BEZ235, in three out of than PLX4720-BEZ235 at inhibiting p-ERK in Me13 nine experimental conditions where both AZD6244 and after O/N treatment, and AZD6244-BEZ235 was the PLX4720 were present at equivalent doses (Figure 4A, most effective treatment as suppressing p-ERK in statistical comparisons highlighted by blue lines and Me6 (Figure 5). AZD6244-BEZ235 was also the most asterisks). effective treatment at inhibiting p-AKT in Me13 and In addition to a combination index matrix, the drug Me6 (Figure 5). All three treatments showed similar interaction analysis yielded also a 1,764-point Fraction inhibitory activity on p-S6 (Figure 5). Enhanced p-ERK Affected (FA) matrix (Supplementary Figure 3A, 3B inhibition by AZD6244-BEZ235 and AZD6244-AZD8055 for color coding of FA values and Supplementary Table compared to PLX4720-BEZ235 were observed also 3 for primary data). Analysis of the FA matrix led to on Me79, a strongly cross-resistant cell line in group1 two main conclusions. First, combinatorial treatment (Figure 5) as well as on Me71, a cell line from group 3 with AZD6244-BEZ235 or PLX4720-BEZ235, but to a susceptible to all inhibitors (Supplementary Figure 6). much lesser extent with AZD6244-AZD8055, allowed to The enhanced suppression of p-ERK by AZD6244- achieve FA values > 0.50, or even > 0.70, on most cross- BEZ235 and AZD6244-AZD8055 treatments, compared resistant cell lines in group 1. Second, and most relevant, to PLX4720-BEZ235, was confirmed even after a shorter in each of the 9 matched drug combinations (identified by (4 h) treatment on melanoma cell lines from group 1 matched symbols at bottom of Supplementary Figure 3A) (Me13 and Me79, Supplementary Figure 7). At 4 h of where AZD6244 and PLX4720 were used at equivalent treatment, AZD6244-BEZ235 and PLX4720-BEZ235 doses, AZD6244-BE235 induced significantly higher FA were effective in suppressing p-AKT on Me13 and Me79, values compared to PLX4720-BEZ235 on the 21 cell lines while inhibition of pS6 by the combinatorial treatments in group 1, characterized by strong primary resistance to was strongly reduced on Me13 and not observed on Me79 PLX4720 and by frequent cross-resistance (Supplementary compared to the effects detected on the same cell lines after Figure 4A, 4B for statistical analysis). O/N treatment (see Supplementary Figure 7 vs. Figure 5). We then compared anti-tumor activity in-vivo of Enhanced inhibition of p-ERK and p-AKT by AZD6244- AZD6244-BEZ235 and PLX4720-BEZ235 combinatorial BEZ235 compared to PLX4720-BEZ235, and to control treatments. To this end, we established a SCID mouse animals treated with vehicle, was also observed in-vivo, model based on s.c. xenograft of a cell line from group 1 as indicated by a reduced staining for p-ERK and p-AKT (Me13). This cell line was characterized by strong primary in melanoma cells from neoplastic nodules removed after

resistance to PLX4720 (IC50 > 1 μM) and by intermediate the last administration of inhibitors (Figure 6A, 6B and

resistance to AZD6244 and BEZ235 (IC50 > 0.1 μM). Both Supplementary Figure 8 for quantitative analysis). treatments exerted a significant inhibitory effect on tumor Taken together, these results indicated that the asso growth, compared to control animals treated with vehicle, ciation of MEK1/2 and PI3K/mTOR inhibitors induces but the AZD6244-BEZ235 association was signifi a more effective inhibition of p-ERK at both early and cantly more effective than PLX4720-BEZ235 treatment late time points compared to BRAF and PI3K/mTOR dual (Supplementary Figure 5). Moreover, no evident signs of blockade, in PLX4720-resistant melanoma cells and even treatment-related toxicity were observed (data not shown). in cell lines with a cross-resistant phenotype. Taken together, the in-vitro and in-vivo experiments indicated that co-targeting of MEK1/2 and PI3K/mTOR Enhanced apoptotic response in melanoma cells has improved anti-tumor activity compared to co-targeting with a cross-resistant phenotype by co-targeting of mutant BRAF and PI3K/mTOR even in melanoma cells of MEK1/2 and PI3K/mTOR with an intrinsic cross-resistant phenotype. The most significant effect of AZD6244-BEZ235, MEK1/2 and PI3K/mTOR co-targeting is more compared to PLX4720-BEZ235 and to AZD6244-AZD8055, effective than BRAF and PI3K/mTOR dual was a reciprocal shift in the sub-G1 and G1 fractions, while blockade in inhibition of signaling molecule S and G2M phases of the cell cycle were not differentially phosphorylation in PLX4720-resistant and in affected, as documented by DNA content analysis cross-resistant melanoma cell lines (Supplementary Figure 9) in representative melanoma cell lines from the three PLX4720 susceptibility groups. This Changes in the phosphorylation status of ERK, AKT result was consistent with enhanced induction of cell death and S6 by the combinatorial treatments were assessed in by AZD6244-BEZ235. By annexin-V/PI staining assays we three cell lines from group 1 (Me13, Me6 and Me79). compared extent of apoptosis induced by single inhibitors and by the three combinatorial treatments on representative Me13 and Me6 have a higher IC50 for PLX4720 than for AZD6244, therefore PLX4720 was used at a higher cell lines from the three groups. To this end, inhibitors www.impactjournals.com/oncotarget 3953 Oncotarget Figure 5: Inhibition of p-ERK and p-AKT by combinatorial treatments in melanoma cells with intrinsic resistance to PLX4720. Three melanoma cell lines from group 1 (Me13, Me6 and Me79) were treated O/N with AZD6244, BEZ235, PLX4720, AZD8055, or the indicated combinations, and then assessed by western blot for inhibition of relevant signaling molecules. Drug doses for Me13 and Me6 were: AZD6244: 0.1 μM; BEZ235: 0.1 μM; PLX4720: 0.5 μM; AZD8055: 0.1 μM. Drug doses for Me79 were 0.5 μM for all inhibitors. www.impactjournals.com/oncotarget 3954 Oncotarget Figure 6: Inhibition of p-ERK and p-AKT in-vivo by combinatorial treatments in a PLX4720-resistant cell line. Immunohistochemistry analysis by staining with anti-p-ERK (A) and anti-pAKT (B) antibodies of tumor nodules (images of nodules from two animals are shown for each signaling molecule) removed after the last administration of inhibitors (day 31), from control mice (vehicle) and from mice treated with the association of AZD6244-BEZ235 or of PLX4720-BEZ235, as described in the legend to Supplementary Figure 5. Insets, higher magnification of a representative area of each panel highlighting the extent of staining for p-ERK and p-AKT in melanoma cells. Original magnification, 20x.

Figure 7: Combinatorial treatments promote melanoma apoptosis. Seven melanoma cell lines (representative of the three susceptibility groups) were treated with AZD6244, BEZ235, PLX4720, AZD8055 or the indicated combinations of inhibitors for 72 h, and then apoptosis was assessed by Annexin-V/PI assay. Each histogram is the sum of early (annexin-V+/PI–, grey) and late (annexin-V+/ PI+, white) apoptosis values. Extent of primary necrosis (% Annexin-V-/PI+ cells) was always < 5% in control cells and in any of the treatments. Concentrations of inhibitors: Group 1 (Me6 and Me13): AZD6244 0.1 μM, BEZ235 0.1 μM, PLX4720 0.5 μM, AZD8055 0.3 μM; Group 1 (Me79): 0.5 μM for all inhibitors; Group 2 (Me63, Me17, Me45): AZD6244 0.05 μM, BEZ235 0.1 μM, PLX4720 0.1 μM, AZD8055 0.1 μM; Group 3 (Me 71): AZD6244 0.05 μM, BEZ235 0.1 μM, PLX4720 0.1 μM, AZD8055 0.02 μM. Statistical analysis by ANOVA and SNK test. ***p < 0.001. www.impactjournals.com/oncotarget 3955 Oncotarget doses were tailored to each melanoma group based on the in group 1 (Figure 7). Further apoptosis assays were carried

different IC50 values (see legend, Figure 7). By annexin-V/ out by comparing the three combinatorial treatments on all PI stainings we compared the extent of apoptosis induced 49 melanoma cell lines. These experiments confirmed that by the three combinatorial treatments. AZD6244-BEZ235 AZD6244-BEZ235 was more effective than PLX4720- was found to be the most effective association, not only BEZ235 against the cell lines in group 1 and group 2 against representative cell lines from groups 2 and 3 (Supplementary Figure 10A). (Figure 7), but even against melanoma cells with the strong By protein arrays experiments we found that (Me79) or intermediate (Me13) cross-resistant phenotype AZD6244-BEZ235 was more effective than PLX4720-

Figure 8: Modulation of pro- and anti-apoptotic proteins by combinatorial treatments. Expression of apoptosis-related proteins, by protein array screening, in Me13, Me41 and Me71 cells (representing each of three different susceptibility groups defined in Figure 1) at 48 hours of treatment with or without AZD6244 (A62), BEZ235 (BEZ), PLX4720 (PLX), AZD8055 (A80) and the combinations

AZD6244-BEZ235 (A + B), PLX4720-BEZ235 (P + B) and AZD6244-AZD8055 (A + A). Inhibitors concentrations, tailored to the IC50 values of the three susceptibility groups, were: Group 1 (Me13): AZD6244 0.1 μM, BEZ235 0.1 μM, PLX4720 0.5 μM, AZD8055 0.3 μM; Group 2 (Me41): AZD6244 0.05 μM, BEZ235 0.1 μM, PLX4720 0.1 μM, AZD8055 0.1 μM; Group 3 (Me 71): AZD6244 0.05 μM, BEZ235 0.1 μM, PLX4720 0.1 μM, AZD8055 0.02 μM. Statistical analysis by ANOVA and SNK test. ***p < 0.001; **p < 0.01; *p < 0.05. www.impactjournals.com/oncotarget 3956 Oncotarget BEZ235 at enhancing expression of the Bax protein, a pro- enhanced activation of effector caspases [25] on most cell apoptotic Bcl-2 family member, and at downregulating lines, compared to the other associations. In agreement, the inhibitor of apoptosis c-IAP1 in melanoma cells from AZD6244-BEZ235 combination was more effective than the three groups and of livin/ML-IAP in melanoma cells PLX4720-BEZ235 and AZD6244-AZD8055 associations from group 1 and 2 (Figure 8). By staining Me6 melanoma in inducing caspase 3/7 enzymatic activation in assays cells with the 6A7 antibody, recognizing conformational carried out on all 49 melanoma cell lines (Supplementary changes of Bax, one of the early steps in the induction Figure 10B). Further assays looking at the generation of apoptosis, we found a significantly increased fraction of active, cleaved caspase-3, indicated that AZD6244- of Bax+ melanoma cells upon treatment for 24 h with BEZ235 was the most effective treatment, compared AZD6344-BEZ235 compared to PLX4720-BEZ235 and to PLX4720-BEZ235 and to AZD6244-AZD8055 in to AZD6244-AZD8055 (Figure 9A, 9B). Silencing of melanoma cells from group 3 (Me71) and from group 1 Bax in Me6, by siRNA (Figure 9C), significantly reduced (Me6, Me13 and Me79, Supplementary Figure 11A, 11B). activation of caspase-3 and (Figure 9D, 9E) and apoptosis Assessment of neoplastic nodules removed after the (Figure 9F) upon treatment with AZD6344-BEZ235 or last administration of inhibitors from mice receiving PLX4720-BEZ235. the combinatorial treatments indicated an enhanced On the basis of these results, we then tested whether activation of caspase-3, associated with induction of co-targeting of MEK1/2 and PI3K/mTOR could lead to apoptosis (TUNEL+ cells), in tumors from mice treated

Figure 9: Role of Bax upregulation, by combinatorial treatments, in caspase-3 activation and melanoma apoptosis. (A, B) Me6 melanoma cells were treated for 24 h with the indicated associations of inhibitors and then stained with the Bax conformation- specific mAb 6A7. Inhibitors doses as in Figure 7. (B) Mean of three experiments. (C) Western blot analysis for Bax at 48 h and 96 h after transfection with Bax-specific Smart Pool siRNA or with the corresponding negative control siRNA. (D–F) Me6 cells, transfected with Bax-specific Smart Pool siRNA or with the negative control siRNA, were treated at 48 h with the AZD6244-BEZ235 or PLX4720-BEZ235 associations and analyzed at 96 h for active, cleaved caspase-3 (D, E) or for apoptosis (F). Numbers in each panel in A, D: % positive cells, markers set based on staining with secondary antibody only (A) or isotype control (D). Statistical analysis by ANOVA and SNK test (B) or two-way ANOVA and Bonferroni test (E, F). **p < 0.01, ***p < 0.001. www.impactjournals.com/oncotarget 3957 Oncotarget with AZD6244-BEZ235 compared to PLX4720-BEZ25 stacked bar charts in Supplementary Figure 15). Thus, (Supplementary Figure 12). upon combinatorial treatment with AZD6244-BEZ235, In-vitro, a pan-caspase inhibitor z-VAD-fmk, but not the two top canonical pathways affected were the “PI3K/ the negative control z-FA-fmk, significantly reduced Me6 AKT signaling”, and the “PTEN signaling”. In contrast, cells apoptosis promoted by the combinatorial treatments, upon melanoma treatment with PLX4720-BEZ235 or with indicating that melanoma apoptosis was caspase- AZD6244-AZD8055 these two pathways ranked third and dependent (Supplementary Figure 13A, 13B). second, or 15th and sixth, respectively (Supplementary Taken together, these results suggest that co-targeting Figure 15). of MEK1/2 and PI3K/mTOR, compared to BRAF and PI3K/ By looking at genes belonging to three canonical mTOR dual blockade, is a more effective approach to rescue pathways (ERK/MAPK, PI3K/AKT, and “prostate susceptibility to caspase-dependent apoptosis in melanoma cancer signaling”) specific differences in the modulatory cells with intrinsic cross-resistant phenotype. effects induced by the three combinatorial treatments were identified (highlighted by red arrows in Figure 10 Selective modulation of genes in the ERK/MAPK and Supplementary Figure 16A–16C). This comparison and PI3K/AKT canonical pathways by MEK1/2 indicated that AZD6244-BEZ235 induced a selective and PI3K/mTOR dual blockade compared to downmodulation of several genes, including c-FOS BRAF and PI3K/mTOR co-targeting (Figure 10A), recently involved in melanoma resistance to MAPK inhibition [13], and p90RSK encoding for Whole genome gene expression analysis was carried proteins that phosphorylate CREB transcription factors out in the PLX4720-resistant Me13 cell line treated with [27–28], also involved in melanoma resistance to AZD6244-BEZ235, or PLX4720-BEZ235 or AZD6244- MAPK inhibition [13]. The preferential downmodulation AZD8055 associations, to identify combination-specific of c-FOS by AZD6244-BEZ235, compared to the effects. To improve the statistical analysis of the results, effects of the other associations of inhibitors, was three independent biological replicates for each treatment confirmed by qPCR (Supplementary Figure 17A). were analyzed. Moreover, drug doses were chosen AZD6244-BEZ235, but not PLX4720-BEZ235, downmodulated the anti-apoptotic gene Bcl-2 and according to the different IC50 values; thus, PLX4720 was used at 0.5 μM while AZD6244 was used at 0.1 μM the mTOR interactor 4EBP1 (Figure 10B), as well as (see Materials and methods for details). Significantly β catenin, a gene that impairs T cell-mediated immune modulated genes by each combination, identified by class response in melanoma [29], and LEF-1 [30] a β catenin comparison through BRB Array Tools, were subjected interacting partner (Supplementary Figure 16B, 16C). to downstream effect analysis through IPA software. The latter two genes are upstream of cyclin D1, a gene “Cell death and survival” was the top biological function being more strongly inhibited (FC = −3.85) by AZD6244- affected by the three combinatorial treatments, but the most BEZ235, compared to PLX4720-BEZ235 (FC = −2.17). significantP value for association with such function was β catenin downmodulation by AZD6244-BEZ235 observed for the AZD6244-BEZ235 treatment (data not treatment was confirmed by qPCR (data not shown). By shown). Based on Z scores > 2 or < -2, the function “cell immunohistochemistry in neoplastic nodules removed death” was predicted to be increased, while the function after the last administration of inhibitors, a reduced “proliferation” was predicted to be decreased by the three staining for β catenin was observed in melanoma cells combinatorial treatments (see Supplementary Table 4 for from animals treated with AZD6244-BEZ235 compared results on AZD6244-BEZ235 treatment). By Edwards- to animals receiving vehicle or the PLX4720-BEZ235 VENN diagram analysis [26] combination-specific gene combination (Supplementary Figure 17B). expression changes were identified (Supplementary Collectively, this evidence indicates that the Figure 14, underlined values) in addition to gene AZD6244-BEZ235 association has an enhanced effect expression changes shared by two or even three different on the biological function “cell death and survival” and combinatorial treatments (Supplementary Figure 14, a selective modulatory effect on genes that play a role in boxed values). Thus, AZD6244-BEZ235 upregulated melanoma resistance to target-specific inhibitors and in a set of 79 genes and downregulated a different set of suppression of anti-tumor immunity. 83 genes, not significantly affected by the other two combinatorial treatments. By canonical pathway analysis DISCUSSION by IPA (Supplementary Figure 15), the specificity of each combinatorial treatment was revealed by two parameters: Intrinsic resistance to BRAF inhibitors prevents a) the different ranking of each of the top 15 pathways ~20% of melanoma patients with BRAF-mutant tumors (based on the P value of the association of the genes with from achieving clinical benefit from this type of target the pathway) and, b) the differences, among the three therapy and represents a major clinical issue [31, for combinatorial treatments, in the fraction of significantly review]. In this subset of patients, combination treatments modulated genes belonging to each pathway (plotted as based on co-targeting of different oncogenic pathways www.impactjournals.com/oncotarget 3958 Oncotarget Figure 10: Selective modulation of genes in the ERK/MAPK and PI3K/AKT canonical pathways by AZD6244-BEZ235. (A, B) genes modulated in Me13 cells by AZD6244-BEZ235 treatment and belonging to the ERK/MAPK canonical pathway (A) and the PI3K/AKT canonical pathway (B). Upregulated genes are shown in red, and downregulated genes in green. Red arrows: genes differently affected by AZD6244-BEZ235 compared to PLX4720-BEZ235 and to AZD6244-AZD8055 treatments (see Supplementary Figure 16 for genes modulated by the latter two treatments). www.impactjournals.com/oncotarget 3959 Oncotarget have been suggested as a potentially effective approaches caspase-3 activation and melanoma apoptosis. Further [32], but the best combinatorial association has not been assays with a pan-caspase inhibitor also indicated that identified yet. Clinical studies of combinatorial treatment melanoma apoptosis was significantly inhibited by a with MEK and PI3K/mTOR inhibitors are ongoing in pan-caspase inhibitor, suggesting that one of the relevant different solid tumors, including melanoma (source: effects of the combinatorial treatments is to rescue www.clinicaltrials.gov) and results of published trials [33] susceptibility of melanoma to caspase-dependent cell suggest that dual targeting of these pathways has clinical death. activity, although toxicity issues have been identified. Analysis for combination-specific effects on gene Moreover, pre-clinical studies, in different tumors suggest expression, provided evidence for selective modulation that co-targeting of MAPK and PI3K/mTOR pathways of genes belonging to the ERK/MAPK and PI3K/AKT may be a potentially promising strategy, as shown by canonical pathways by the AZD6244-BEZ235, compared results obtained in gefitinib-resistant NSCLC cells [34], to PLX4720-BEZ235 and to AZD6244-AZD8055. The in pancreatic cancer cells [35], in rhabdomyosarcoma cells inhibitory effects on c-FOS, and on genes (p90RSK) [36], in NRAS mutant melanoma cells [17] and in BRAF encoding proteins that affect CREB phosphorylation, mutant melanoma cells with acquired resistance mediated suggest that AZD6244-BEZ235 is a potentially effective by upregulation of PDGFRβ [16]. approach to overcome recently described mechanisms In this study we provide novel preclinical evidence mediating melanoma resistance to MAPK inhibition indicating that primary resistance to BRAF inhibition [13]. In agreement with this interpretation, caspase 3/7 is not only frequently associated with cross-resistance activation assays and cell death assays indicated that to MEK1/2 inhibitors, but even with cross-resistance to AZD6244-BEZ235 was significantly more effective in PI3K/mTOR inhibitors. Based on this evidence, we the promotion of apoptosis on melanoma cell lines of tested whether a combinatorial treatment approach could all susceptibility groups. Thus, these results support the be effective in the subset of melanoma cells with the notion that rescuing susceptibility to apoptosis is a major intrinsic cross-resistant phenotype. The results indicated mechanism of action of effective co-targeting strategies that the most significant synergistic effects (as determined in melanoma, as reported previously by us [18, 37–38], by combination index values) could be achieved by co- and other groups in melanoma [16] and other tumors [35]. targeting of MEK1/2 and PI3K/mTOR. Co-targeting The selective downmodulation of β catenin, by of MEK1/2 and PI3K/mTOR was significantly more AZD6244-BEZ235 compared to PLX4720-BEZ235, effective than BRAF and PI3K/mTOR dual blockade suggests that dual blockade of the MAPK and PI3K/ not only in-vitro, but also in-vivo, against xenografts mTOR pathways could suppress a recently discovered from a cell line with intrinsic resistance to the BRAF mechanism that melanoma cells exploit to suppress inhibitor PLX4720. The anti-melanoma efficacy of the development of T cell mediated anti-tumor response AZD6244-BEZ235 treatment was associated with strong [29] and provides a further rationale for the association inhibition of key signaling molecules, as documented by of target therapy, based on dual pathway inhibition, with modulation of p-ERK and of p-AKT, both in-vitro and immune checkpoint blockade, an approach that has greatly in-vivo. Interestingly, in melanoma cells with acquired improved the management of advanced disease [39]. resistance to BRAF inhibition, Shi et al. [16] found that Analysis of responsiveness to BRAF, MEK1/2 and PLX4720, alone or in combination with BEZ235 or PI3K/mTOR inhibitors in a few short term melanoma AZD8055, induced early and delayed p-ERK recovery, cell cultures, from patients subsequently treated with while AZD6244 in association with BEZ235 or AZD8055, target therapy, suggested that drug susceptibility data strongly reduced such effect. may predict response or resistance to treatment. Clearly, Assessment of the mechanism(s) involved in the this hypothesis needs confirmation in a larger set of enhanced anti-melanoma efficacy by co-targeting of patients, but the testing of freshly isolated melanoma cells MEK1/2 and PI3K/mTOR, compared to mutant BRAF from surgical samples, for responsiveness to available and PI3K/mTOR dual inhibition, provided evidence inhibitors, is feasible in principle in advanced melanoma for a more effective induction of apoptosis, not only and could provide a valuable evidence to inform in-vitro, but also in-vivo and even in cross-resistant subsequent clinical decisions. cell lines. Treatment by AZD6244-BEZ235 was associated with enhanced modulation of pro- and anti- MATERIALS AND METHODS apoptotic molecules, compared to PLX4720-BEZ235 and to AZD6244-AZD8055. Further assessment of the Ethics statement mechanism of melanoma apoptosis, after dual pathway co-targeting, indicated that AZD6244-BEZ235 was the This investigation has been conducted in accordance most effective combination at inducing activation of Bax with the ethical standards and according to the Declaration and of caspase-3, the latter effect being observed both of Helsinki and according to national and international in-vitro and in-vivo. Silencing of Bax reduced both guidelines and has been approved by the independent www.impactjournals.com/oncotarget 3960 Oncotarget ethical committee of our Institute. In-vivo experiments followed by Image J quantification of images. Results in SCID mice were performed according to the Italian of clonogenic assays were expressed as % inhibition laws (D.L. 116/92 and after additions), after approval of melanoma growth. For drug interaction analysis the by the institutional Ethical Committee for Animal Chou and Talalay method was used [24]. To this end, Experimentation of our Institute and by the Italian MTT assays were set up testing three combinations of Ministry of Health (Project INT_17/2011). inhibitors (AZD6244-BEZ235, PLX4720-BEZ235 and AZD6244-AZD8055). For each combinatorial treatment, Cell lines and short-term melanoma cell cultures twelve different combinations of doses were used. Data were then analyzed to obtain Combination index (CI) and Forty-nine BRAF-mutant melanoma cell lines Fraction Affected (FA) values by the CompuSyn software were established as described [18, 37–38, 40–41] from (ComboSyn). surgical specimens of American Joint Committee on Cancer (AJCC) stage IIIc and IV melanoma patients not previously subjected to target therapy. An independent Antibodies and western blot analysis panel of 33 short-term melanoma cell cultures was Western blot analysis was carried out by the following generated as described [40–41] from surgical samples of antibodies specific for: AKT, p-AKT (Ser473), ERK 1/2, AJCC stage IIIc and IV BRAF mutant (n = 23) or wild p-ERK1/2 (Thr202/Tyr204), S6, p-S6 (Ser235/236), PTEN type (n = 10) patients not previously subjected to target therapy. Short-term melanoma cell cultures were used and Bax (Cell Signaling); MDM4 (Bethyl Laboratories), between the third and fifth in-vitro passage. All patients MDM2 (Santa Cruz), β-actin (Sigma-Aldrich), α-tubulin were admitted to Fondazione IRCCS Istituto Nazionale (Calbiochem) and Vinculin (Sigma-Aldrich). SDS-PAGE dei Tumori, Milan and all the lesions were histologically was performed using 30 μg of protein samples on 4%–12% confirmed to be cutaneous malignant melanomas. Informed NuPAGE Bis-Tris polyacrylamide gels (Invitrogen Life consent was obtained from patients. Molecular and Technologies), as described [18, 37–38]. Development biological characterization of the cell lines and methods was performed by the chemiluminescence method with the for identification of mutations in BRAF gene have been ECL Western Blotting Detection System (GE Healthcare) reported previously [18, 37–38, 41–42]. or Luminata Crescendo (Millipore). Basal/constitutive phosphoprotein characterization was carried out after Treatment of melanoma cells with inhibitors and O/N culture without FCS. Modulation of expression of drug interaction analysis phosphoproteins, by inhibitors, was assessed in melanoma cells treated either for 4 hr or O/N with MAPK and PI3K/ Two days before treatment, melanoma cells were mTOR pathways inhibitors, alone or in combinations, seeded in 96-wells flat bottom plates in RPMI 1640 with 2% FCS. Protein quantification was performed by (BioWhittaker) supplemented with 2% fetal calf serum densitometric analysis with the Quantity One software (FCS) without antibiotics. For IC50 determination, (BioRad Laboratories). treatments were in quadruplicate with AZD6244 (MEK1/2 inhibitor, SelleckChem), PLX4720 (BRAFV600E In-vivo evaluation of anti-melanoma activity of inhibitor, SelleckChem), BEZ235 (dual PI3K and mTOR combinatorial treatments inhibitor, SelleckChem), AZD8055 (dual mTORC1/2 inhibitor, SelleckChem). Stocks and dilutions of inhibitors were Female SCID mice, 8–10 weeks old (Charles River done in DMSO. Twelve concentrations of each inhibitor Laboratories) were provided with food and water ad ranging from 1 nM to 10 μM at final FCS concentration libitum. Melanoma cells (Me13), harvested in exponential of 1% were used. Cultures were evaluated at 72 hours as growth phase, were injected s.c. (5x106) in the left flank described [18, 37–38] by the 3-(4, 5) dimethylthiazol-2, of each mouse. When tumors became palpable, mice 5-diphenyltetrazolium bromide (MTT) assay. IC50 values were obtained through nonlinear regression analysis (by were randomized into three groups (7 animals/group) PRISM software, Graphpad) of dose-response curves by and animals received either vehicle, the association of a log (inhibitor) vs. response, variable slope equation. AZD6244 (10 mg/kg) and BEZ235 (20 mg/Kg) or of Clonogenic assays were performed by seeding melanoma PLX4720 (10 mg/Kg) and BEZ235 (20 mg/Kg), 5 days cells at single-cell density in 6-well plates. Treatments per week for three consecutive weeks by oral gavage. with AZD6244, BEZ235, PLX4720, and AZD8055 Mice were monitored daily for signs of toxicity and were (0.1–0.5 μM), either alone or in combination, vs. DMSO weighed twice weekly. Tumor size was regularly evaluated were done every 72 h. After 12 days, the supernatant was by measuring the orthogonal diameters (d and D) and discarded, plates were washed with HBSS, fixed with calculating the volumes with the following formula: 4/3π methanol and then stained with Giemsa (Sigma Aldrich) [(d2D)/2].

www.impactjournals.com/oncotarget 3961 Oncotarget Flow cytometry analysis containing four oligos specific for Bax (#J-003308–11, –12, –13, –14, Dharmacon) or corresponding negative Melanoma apoptosis was assessed after staining control (#D-001320–10–05, Lincode non targeting with APC-conjugated Annexin V (BD Pharmingen) and pool, Dharmacon). Oligos were used at 10 nm final propidium iodide (PI; BD Biosciences) as described [37– concentration according to Lipofectamine RNAiMAX 38]. DNA content analysis was carried out after staining guidelines (Thermo Fisher Scientific). Silencing was with propidium iodide (Sigma-Aldrich) as described [37]. checked by Western blot analysis at 48 h and at 96 h. At Staining with purified mouse Bax conformation-specific 48 h after transfection with Bax siRNA smart pool, or with antibody 6A7 or with rabbit FITC-anti-active caspase-3 negative control oligos, cells were treated with AZD6244 antibody (BD Pharmingen) was carried out on cells plus BEZ235 (0.1 μM + 0.1 μM), or with PLX4720 plus permeabilized with CytoFix/CytoPerm (BD Pharmingen). BEZ235 (0.1 μM + 0.5 μM) and analyzed after 48 h by Secondary and control antibodies were, respectively, staining for Annexin-V/Propidium Iodide or for active FITC-labeled goat anti-mouse (Jackson Immunoresearch caspase-3. Laboratories) and FITC-rabbit IgG isotype control. In some experiments, melanoma cells were pre-incubated Immunohistochemistry with general caspase inhibitor z-VAD-fmk or control z-FA-fmk (BD Pharminge) at 5 μM for 1 h at 37°C before Immunohistochemistry was performed with treatment with drugs. All experiments were carried out formalin-fixed, paraffin-embedded tissues as described with a FACSCalibur flow cytometer (BD Biosciences) and [40]. SCID mice bearing s.c. Me13 xenografts were analyzed by the FlowJo software (Tree Star). treated with AZD6244-BEZ25 or PLX4720-BEZ235 combinations as described for the tumor growth Cell death and caspase 3/7 activity assays by inhibition assays. Neoplastic nodules were removed Muse cell analyzer after the last administration of inhibitors (day 31) and were characterized by staining with antibodies to p-ERK In some experiments, apoptosis and caspase 3/7 (Thr202/Tyr204, Cell Signaling), p-AKT (Ser473, Cell activation were assessed by the Muse™ Cell Analyzer Signaling), β catenin (BD Transduction Laboratories), (Merck Millipore). For apoptosis determination, cells were or cleaved caspase-3 (Cell Signaling). The extent of collected and resuspended according to manufacturer’s apoptosis in neoplastic nodules was evaluated by TUNEL instructions with the working solution of the Muse™ staining (Roche). Cytospin preparations of melanoma Annexin V & Dead Cell Kit (Merck Millipore) for cell lines were processed and stained for AXL (R & D 20 minutes at room temperature. For caspase 3/7 activity Systems) and for MITF (Dako) as described [43]. Images cells were collected, resuspended in 1X Assay Buffer BA, were acquired at 20x with an Axiovert 100 microscope and incubated with Muse™ Caspase-3/7 Assay Kit (Merck (Zeiss) equipped with a digital camera (AxioCam MrC5, Millipore) reagent working solution for 30 minutes in the Zeiss). For image acquisition, all main microscope and ™ 37°C incubator with 5% CO2. After incubation, Muse digital camera operative settings, including exposure time, Caspase 7-AAD working solution was added for 5 minutes were kept constant. Immunohistochemistry images were at room temperature. Data were analyzed by Muse 1.4 quantified using ImageJ. Analysis Software. Genome-wide expression profiling of melanoma Apoptosis antibody array cells treated with MAPK- and PI3K/mTOR- specific inhibitors Melanoma cells were treated with MAPK and PI3K/mTOR inhibitors for 48 h. The Human Apoptosis Melanoma cells from Me13 cell line were treated with Array Kit (R & D Systems) was used, as described [37] AZD6244 (0.1 μM), BEZ235 (0.1 μM), AZD8055 (0.3 μM) according to manufacturer’s instructions. The intensity of or PLX4720 (0.5 μM), or with the AZD6244-BEZ235, protein signals was quantified by densitometric analysis PLX4720-BEZ235, AZD6244-AZD8055 combinations for with the Quantity One software (BioRad Laboratories). 8 hr. Three biological replicates for each treatment were set After background subtraction, results were expressed up. Total RNA isolation, clean-up, DNase treatment and as the percentage of the mean of the relative positive assessment of RNA integrity and purity were performed controls. as described [18, 38]. Single-color hybridization of RNAs was performed on Illumina Bead Chip HumanHT-12_v4 Bax silencing experiments Microarrays (Illumina). The expression profiles have been deposited in NCBI’s Gene Expression Omnibus (GEO) Transient silencing experiments were carried out with GSE accession number GSE59882. Background with ON-TARGET Plus Human BAX siRNA smart- correction, filtering of data, and quantile normalization were pool (#L-003308–01–0005, Dharmacon GE Healthcare) done using the BeadStudio Illumina software. Identification www.impactjournals.com/oncotarget 3962 Oncotarget of significantly modulated genes was carried out by BRB Fondazione IRCCS Istituto Nazionale dei Tumori, Milan, array tools (vers 4.3.0) developed by Dr. Richard Simon and for gene expression profiling experiments. Amy Peng Lam. Generation of Edwards-VENN diagrams was obtained by VENNTURE software [26]. Downstream GRANT SUPPORT effects analysis and canonical pathway analysis were performed by Ingenuity Pathway analysis, IPA 8.5 (www. Supported by grant #11608 (to A.A.) from Associ ingenuity.com) as described [18]. azione Italiana per la Ricerca sul Cancro (A.I.R.C., Milan).

Real time PCR CONFLICTS OF INTEREST c-Fos (Hs99999140-m1), CTNNB1/β catenin The authors have no conflicts of interest to disclose. (Hs003550489-m1) and, as endogenous control, GAPDH (Hs00266705-g1) TaqMan Gene Expression Assays (Applied Biosystems, Foster City, CA) were used. Total RNA REFERENCES (1 μg) was reverse transcribed with oligo d (t) using a Transcriptor First Strand cDNA Synthesis Kit (Roche, 1. Chapman PB, Hauschild A, Robert C, Haanen JB, Ascierto P, Penzberg, Germany). Preliminary experiments were Larkin J, Dummer R, Garbe C, Testori A, Maio M, Hogg D, Lorigan P, Lebbe C, et al. Improved survival with conducted for the endogenous control using the Ct slope method to ensure that the quality of each complementary vemurafenib in melanoma with BRAF V600E mutation. N DNA and the dynamic range of amplifications were Engl J Med. 2011; 364:2507–2516. comparable [44]. Real-time PCR was then carried out 2. Hauschild A, Grob JJ, Demidov LV, Jouary T, Gutzmer R, with 20 ng input complementary DNA, 1 × TaqMan Gene Millward M, Rutkowski P, Blank CU, Miller WH Jr, Expression Master Mix and TaqMan Gene Expression Kaempgen E, Martín-Algarra S, Karaszewska B, Mauch C, Assays on a ABI PRISM 7900 HT thermal cycler (Applied et al. Dabrafenib in BRAF-mutated metastatic melanoma: a Biosystems). Data were analyzed using ABI PRISM multicentre, open label, phase 3 randomized controlled trial. Sequence Detection Software version 2.2.2 (Applied Lancet. 2012; 380:358–365. Biosystems). Relative expression was determined on 3. Sosman JA, Kim KB, Schuchter L, Gonzalez R, Pavlick AC, triplicate reactions using the formula 2−ΔCt, reflecting Weber JS, McArthur GA, Hutson TE, Moschos SJ, Flaherty KT, target gene expression normalized to endogenous control Hersey P, Kefford R, Lawrence D, et al. Survival in levels [44]. BRAFV600E-mutant advanced melanoma treated with Vemurafenib. N Engl J Med. 2012; 366:707–14. Statistical analysis 4. Kugel CH 3rd, Aplin AE. Adaptive resistance to RAF inhibitors in melanoma. Pigment Cell Mel Res. 2014; IC50 values were clustered by Cluster 3.0 software 27:1032–1038. and clustering was visualized by Java TreeView. 5. Aplin AE, Kaplan FM, Shao Y. Mechanisms of resistance Correlation of susceptibility profiles to different inhibitors to RAF inhibitors in melanoma. J Invest Dermatol. 2011; was tested by Spearman correlation analysis. Comparison 131:1817–1820. of IC values, of combination index (CI) values and 50 6. Falchook GS, Lewis KD, Infante JR, Gordon MS, of fraction affected (FA) data, in different groups of Vogelzang NJ, DeMarini DJ, Sun P, Moy C, Szabo SA, melanoma cell lines, was done by Kruskal-Wallis and Roadcap LT, Peddareddigari VG, Lebowitz PF, Le NT, et al. Dunn multiple comparison test. Analysis of FA values Activity of the oral MEK inhibitor trametinib in patients by combinations of inhibitors in the same group of with advanced melanoma: a phase 1 dose-escalation trial. melanoma cell lines, or in all cell lines, when equivalent Lancet Oncol. 2012; 13:782–789. doses of different inhibitors were used, was carried out by Wilcoxon matched pair test. Significance of different 7. Catalanotti F, Solit DB, Pulitzer MP, Berger MF, Scott SN, treatments on melanoma apoptosis, caspase activation, Iyriboz T, Lacouture ME, Panageas KS, Wolchok JD, modulation of apoptosis-related molecules was assessed Carvajal RD, Schwartz GK, Rosen N, Chapman PB. by ANOVA, followed by Student-Newman-Keul (SNK) Phase II trial of MEK inhibitor selumetinib (AZD6244, multiple comparison test. Analysis of the antitumor ARRY-142886) in patients with BRAFV600E/K-mutated activity of different treatments in-vivo was carried out by melanoma. Clin Cancer Res. 2013; 19:2257–2264. mixed effects model ANOVA [45] by XLSTAT software 8. Long GV, Stroyakovskiy D, Gogas H, Levchenko E, de (Xlstat). Braud F, Larkin J, Garbe C, Jouary T, Hauschild A, Grob JJ, Chiarion Sileni V, Lebbe C, Mandalà M, et al. Combined ACKNOWLEDGMENTS BRAF and MEK inhibition versus BRAF inhibition alone in melanoma. N Engl J Med. 2014; 371:1877–1888. The authors wish to thank Dr. L. De Cecco and 9. Larkin J, Ascierto PA, Dréno B, Atkinson V, Liszkay G, Mr. E. Marchesi of the Functional Genomics Facility of Maio M, Mandalà M, Demidov L, Stroyakovskiy D, www.impactjournals.com/oncotarget 3963 Oncotarget Thomas L, de la Cruz-Merino L, Dutriaux C, Garbe C, et al. BRAFV600E inhibitor, displays potent antitumor activity Combined vemurafenib and cobimetinib in BRAF-mutated in preclinical melanoma models. Cancer Res. 2010; 70: melanoma. N Engl J Med. 2014; 371:1867–1876. 5518–5527. 10. Robert C, Karaszewska B, Schachter J, Rutkowski P, 20. Ji Z, Kumar R, Taylor M, Rajadurai A, Marzuka-Alcalá A, Mackiewicz A, Stroiakovski D, Lichinitser M, Dummer R, Chen YE, Njauw CN, Flaherty K, Jönsson G, Tsao H. Grange F, Mortier L, Chiarion-Sileni V, Drucis K, Krajsova I, Vemurafenib synergizes with nutlin-3 to deplete survivin et al. Improved overall survival in melanoma with combined and suppresses melanoma viability and tumor growth. Clin dabrafenib and trametinib. N Engl J Med. 2015; 372:30–39. Cancer Res. 2013; 19:4383–4391. 11. Konieczkowski DJ, Johannessen CM, Abudayyeh O, Kim JW, 21. Paraiso KH, Xiang Y, Rebecca VW, Abel EV, Chen YA, Cooper ZA, Piris A, Frederick DT, Barzily-Rokni M, Munko AC, Wood E, Fedorenko IV, Sondak VK, Straussman R, Haq R, Fisher DE, Mesirov JP, Hahn WC, Anderson AR, Ribas A, Palma MD, Nathanson KL, et al. et al. A melanoma cell state distinction influences sensitivity PTEN loss confers BRAF inhibitor resistance to melanoma to MAPK inhibitors. Cancer Discov. 2014; 4:816–827. cells through the suppression of BIM expression. Cancer 12. Müller J, Krijgsman O, Tsoi J, Robert L, Hugo W, Song C, Res. 2011; 71:2750–2760. Kong X, Possik PA, Cornelissen-Steijger PDM, Geukes 22. Gembarska A, Luciani F, Fedele C, Russell EA, Dewaele M, Foppen MH, Kemper K, Goding CR, McDermott U, et al. Villar S, Zwolinska A, Haupt S, de Lange J, Yip D, Goydos J, Low MITF/AXL ratio predicts early resistance to multiple Haigh JJ, Haupt Y, et al. MDM4 is a key therapeutic target targeted drugs in melanoma. Nat Commun. 2014; 5:5712. in cutaneous melanoma. Nature Med. 2012; 18:1239–1247. 13. Johannessen CM, Johnson LA, Piccioni F, Townes A, 23. Lu M, Breyssens H, Salter V, Zhong S, Hu Y, Baer C, Frederick DT, Donahue MK, Narayan R, Flaherty KT, Ratnayaka I, Sullivan A, Brown NR, Endicott J, Knapp S, Wargo JA, Root DE, Garraway LA. A melanocyte lineage Kessler BM, Middleton MR, et al. Restoring p53 function program confers resistance to MAP kinase pathway in human melanoma cells by inhibiting MDM2 and cyclin inhibition. Nature. 2013; 138:504–8. B1/CDK1-phosphorylated nuclear iASPP. Cancer Cell. 2013; 23:618–33. 14. Atefi M, von Euw E, Attar N, Ng C, Chu C, Guo D, Nazarian R, Chmielowski B, Glaspy JA, Comin-Anduix B, 24. Chou TC. Theoretical basis, experimental design, and Mischel PS, Lo RS, Ribas A. Reversing melanoma cross- computerized simulation of synergism and antagonism resistance to BRAF and MEK inhibitors by co-targeting the in drug combination studies. Pharmacol Rev. 2006; 58: AKT/mTOR pathway. PLOS One. 2011; 6:e28973. 621–81. 15. Roberts PJ, Usary JE, Darr DB, Dillon PM, Pfefferle AD, 25. Tenev T, Zachariou A, Wilson R, Ditzel M, Meier P. IAPs Whittle MC, Duncan JS, Johnson SM, Combest AJ, Jin J, are functionally non-equivalent and regulate effector Zamboni WC, Johnson GL, Perou CM, et al. Combined caspases through distinct mechanisms Nature Cell Biology. PI3K/mTOR and MEK inhibition provides broad antitumor 2004; 7:70–77. activity in faithful murine cancer models. Clin Cancer Res. 26. Martin B, Chadwick W, Yi T, Park SS, Lu D, Ni B, 2102; 18:5290–5303. Gadkaree S, Farhang K, Becker KG, Maudsley S. 16. Shi H, Kong X, Ribas A, Lo RS. Combinatorial treatments VENNTURE–a novel Venn diagram investigational tool that overcome PDGFRβ-driven resistance of melanoma for multiple pharmacological dataset analysis. PLOS One. cells to V600E B-RAF inhibition. Cancer Res. 2011; 2012; 7:e36911. 71:5067–74. 27. Arthur JS. MSK activation and physiological roles. Front 17. Posch C, Moslehi H, Feeney L, Green GA, Ebaee A, Biosci. 2008; 13:5866–79. Feichtenschlager V, Chong K, Peng L, Dimon MT, Phillips T, 28. Frödin M, Gammeltoft S. Role and regulation of 90 kDa Daud AI, McCalmont TH, LeBoit PE, et al. Combined ribosomal S6 kinase (RSK) in signal transduction. Mol Cell targeting of MEK and PI3K/mTOR effector pathways is Endocrinol. 1999; 151:65–77. necessary to effectively inhibit NRAS mutant melanoma 29. Spranger S, Bao R, Gajewski TF.Melanoma-intrinsic in vitro and in vivo. Proc Natl Acad Sci USA. 2013; 110: β-catenin signalling prevents anti-tumour immunity. Nature. 4015–4020. 2015. doi: 10.1038/nature14404. [Epub ahead of print]. 18. Grazia G, Vegetti C, Benigni F, Penna I, Perotti V, Tassi E, 30. Jamieson C, Sharma M, Henderson BR. Targeting the Bersani I, Nicolini G, Canevari S, Carlo-Stella C, Gianni AM, β-catenin nuclear transport pathway in cancer. Semin Mortarini R, Anichini A. Synergistic anti-tumor activity and Cancer Biol. 2014; 27:20–29. inhibition of angiogenesis by cotargeting of oncogenic and 31. Fedorenko IV, Paraiso KH, Smalley KS. Acquired and death receptor pathways in human melanoma. Cell Death intrinsic BRAF inhibitor resistance in BRAF V600E mutant Dis. 2014; 5:e1434. melanoma. Biochem Pharmacol. 2011; 82:201–9. 19. Yang H, Higgins B, Kolinsky K, Packman K, Go Z, Iyer R, 32. Turajlic S, Furney SJ, Stamp G, Rana S, Ricken G, Oduko Y, Kolis S, Zhao S, Lee R, Grippo JF, Schostack K, Simcox Saturno G, Springer C, Hayes A, Gore M, Larkin J, Marais R. ME, Heimbrook D, et al. RG7204 (PLX4032), a selective Whole-genome sequencing reveals complex mechanisms of www.impactjournals.com/oncotarget 3964 Oncotarget intrinsic resistance to BRAF inhibition. Ann Oncol. 2014; melanoma resistance to antitumor agents that activate the 25:959–67. intrinsic or the extrinsic apoptosis pathways. Clin Cancer 33. Shimizu T, Tolcher AW, Papadopoulos KP, Beeram M, Res. 2012; 18: 3316–27. Rasco DW, Smith LS, Gunn S, Smetzer L, Mays TA, Kaiser B, 39. Shin DS, Ribas A. The evolution of checkpoint blockade Wick MJ, Alvarez C, Cavazos A, et al. The clinical effect as a cancer therapy: what’s here, what’s next? Curr Opin of the dual-targeting strategy involving PI3K/AKT/mTOR Immunol. 2015; 33:23–35. and RAS/MEK/ERK pathways in patients with advanced 40. Anichini A, Mortarini R, Nonaka D, Molla A, Vegetti C, cancer. Clin Cancer Res. 2012; 18:2316–25. Montaldi E, Wang X, Ferrone S. Association of antigen- 34. Qu Y, Wu X, Yin Y, Yang Y, Ma D, Li H. Antitumor activity processing machinery and HLA antigen phenotype of of selective MEK1/2 inhibitor AZD6244 in combination melanoma cells with survival in American Joint Committee with PI3K/mTOR inhibitor BEZ235 in gefitinib resistant on Cancer stage III and IV melanoma patients. Cancer Res. NSCLC xenograft models. J Exp Clin Cancer Res. 2014; 2006; 66:6405–11. 33:52. 41. Daniotti M, Oggionni M, Ranzani T, Vallacchi V, Campi V, Di 35. Zhong H, Sanchez C, Spitrzer D, Plambeck-Suess S, Gibbs J, Stasi D, Torre GD, Perrone F, Luoni C, Suardi S, Frattini M, Hawkins WG, Denardo D, Gao F, Pufahl RA, Lockhart AC, Pilotti S, Anichini A, et al. BRAF alterations are associated Xu M, Linehan D, Weber J, et al. Synergistic effects with complex mutational profiles in malignant melanoma. of concurrent blockade of PI3K and MEK pathways in Oncogene. 2004; 23:5968–77. pancreatic cancer preclinical models. PLOS One. 2013; 42. Sensi M, Nicolini G, Petti C, Bersani I, Lozupone F, 8:e77243. Molla A, Vegetti C, Nonaka D, Mortarini R, Parmiani G, 36. Renshaw J, Taylor KR, Bishop R, Valenti M, De Haven Fais S, Anichini A. Mutually exclusive NRASQ61R and Brandon A, Gowan S, Eccles SA, Ruddle RR, Johnson LD, BRAFV600E mutations at the single-cell level in the same Raynaud FI, Selfe JL, Thway K, Pietsch T, et al. Dual human melanoma. Oncogene. 2006; 25:3357–64. blockade of the PI3K/AKT/mTOR (AZD8055), and RAS/ 43. Mortarini R, Scarito A, Nonaka D, Zanon M, Bersani I, MEK/ERK (AZD6244) pathways synergistically inhibits Montaldi E, Pennacchioli E, Patuzzo R, Santinami M, rhabdomyosarcoma cell growth in vitro and in vivo. Clin Anichini A. Constitutive expression and costimulatory Cancer Res. 2013; 19:5940–51. function of LIGHT/TNFSF14 on human melanoma cells 37. Perotti V, Baldassari P, Bersani I, Molla A, Vegetti C, Tassi E, and melanoma-derived microvesicles. Cancer Res. 2005; Dal Col J, Dolcetti R, Anichini A, Mortarini R. NFATc2 is 65:3428–36. a potential therapeutic target in human melanoma. J Invest 44. Schmittgen TD, Livak KJ. Analyzing real-time PCR data by Dermatol. 2012; 132:2652–60. the comparative C (T) method. Nat Protoc. 2008; 3:1101–8. 38. Tassi E, Zanon M, Vegetti C, Molla A, Bersani I, Perotti V, 45. Liu C, Cripe TP, Kim MO. Statistical issues in longitudinal Pennati M, Zaffaroni N, Milella M, Ferrone S, Carlo-Stella C, data analysis for treatment efficacy studies in the biomedical Gianni AM, Mortarini R, et al. Role of Apollon in human sciences. Mol Ther. 2010; 18:1724–30.